<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    <div class="moz-cite-prefix">On 3/8/2012 2:05 PM, Matthew Knepley
      wrote:<br>
    </div>
    <blockquote
cite="mid:CAMYG4GnSt1vzVr438q+gwUJpGhW4JnaOgahPP-fQdwO6a3P6Ag@mail.gmail.com"
      type="cite">On Fri, Aug 3, 2012 at 4:52 AM, TAY wee-beng <span
        dir="ltr"><<a moz-do-not-send="true"
          href="mailto:zonexo@gmail.com" target="_blank">zonexo@gmail.com</a>></span>
      wrote:<br>
      <div class="gmail_quote">
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div class="im">On 3/8/2012 11:10 AM, thomas wrote:<br>
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex">
              Hi,<br>
              look at the FAQ<br>
              <br>
              <a moz-do-not-send="true"
href="http://www.mcs.anl.gov/petsc/documentation/faq.html#slowerparallel"
                target="_blank">http://www.mcs.anl.gov/petsc/documentation/faq.html#slowerparallel</a><br>
              <br>
              "There must be enough work for each process to overweigh
              the communication time. We recommend an absolute minimum
              of about 10,000 unknowns per process, better is 20,000 or
              more."<br>
            </blockquote>
            <br>
          </div>
          Hi Thomas,<br>
          <br>
          Thanks for the info.<br>
          <br>
          Btw, is there an upper limit? When the unknowns are 100,000 or
          more per process, is it better (as in faster) to use more
          processes instead (e.g. 20,000 per process)?</blockquote>
        <div><br>
        </div>
        <div>That depends on the machine. The lower limit is the minimum
          number of unknowns to cover the latency of network
          communication.</div>
      </div>
    </blockquote>
    <br>
    Thanks for the info!<br>
    <blockquote
cite="mid:CAMYG4GnSt1vzVr438q+gwUJpGhW4JnaOgahPP-fQdwO6a3P6Ag@mail.gmail.com"
      type="cite">
      <div class="gmail_quote">
        <div><br>
        </div>
        <div>   Matt</div>
        <div> </div>
        <blockquote class="gmail_quote" style="margin:0 0 0
          .8ex;border-left:1px #ccc solid;padding-left:1ex">
          <div class="HOEnZb">
            <div class="h5">
              <blockquote class="gmail_quote" style="margin:0 0 0
                .8ex;border-left:1px #ccc solid;padding-left:1ex">
                <br>
                Regards<br>
                Thomas Huxhorn<br>
                <br>
                On 08/03/2012 09:03 AM, TAY wee-beng wrote:<br>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex">
                  Hi,<br>
                  <br>
                  I remembered that in one of the mails, someone
                  mentions the recommended<br>
                  unknowns on each processor when doing parallel
                  solving.<br>
                  <br>
                  However, I can't find that mail. Can someone enlighten
                  me? Also, is this<br>
                  number problem dependent?<br>
                  <br>
                  Btw, I'm using PETSc to solve linear equations using
                  BCGS.<br>
                  <br>
                  Thanks!<br>
                  <br>
                </blockquote>
                <br>
              </blockquote>
              <br>
            </div>
          </div>
        </blockquote>
      </div>
      <br>
      <br clear="all">
      <div><br>
      </div>
      -- <br>
      What most experimenters take for granted before they begin their
      experiments is infinitely more interesting than any results to
      which their experiments lead.<br>
      -- Norbert Wiener<br>
    </blockquote>
    <br>
  </body>
</html>